add checkpoint otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline

.gitattributes

@@ -371,3 +371,14 @@ checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s3_vt_project_a1_L33_bcast/c
 checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s3_vt_project_a1_L33_bcast/checkpoint-525/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s3_vt_project_a1_L33_bcast/checkpoint-5250/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s3_vt_project_a1_L33_bcast/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-12500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-2500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-3750/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-5000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-6250/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-7500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-8750/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/tokenizer.json filter=lfs diff=lfs merge=lfs -text

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/README.md

@@ -0,0 +1,61 @@
+---
+base_model: google/gemma-3-4b-it
+library_name: peft
+model_name: otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline
+tags:
+- base_model:adapter:google/gemma-3-4b-it
+- lora
+- sft
+- transformers
+- trl
+licence: license
+pipeline_tag: text-generation
+---
+
+# Model Card for otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline
+
+This model is a fine-tuned version of [google/gemma-3-4b-it](https://huggingface.co/google/gemma-3-4b-it).
+It has been trained using [TRL](https://github.com/huggingface/trl).
+
+## Quick start
+
+```python
+from transformers import pipeline
+
+question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
+generator = pipeline("text-generation", model="None", device="cuda")
+output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
+print(output["generated_text"])
+```
+
+## Training procedure
+
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/agam-research/huggingface/runs/5gtnw3r4) 
+
+
+This model was trained with SFT.
+
+### Framework versions
+
+- PEFT 0.19.1
+- TRL: 0.28.0
+- Transformers: 4.57.6
+- Pytorch: 2.9.1
+- Datasets: 4.5.0
+- Tokenizers: 0.22.2
+
+## Citations
+
+
+
+Cite TRL as:
+    
+```bibtex
+@software{vonwerra2020trl,
+  title   = {{TRL: Transformers Reinforcement Learning}},
+  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
+  license = {Apache-2.0},
+  url     = {https://github.com/huggingface/trl},
+  year    = {2020}
+}
+```

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-3-4b-it",
+  "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": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "v_proj",
+    "gate_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e396b5c0227fc1a95d5ec2a972d9d020700c32ca7a642ae3bc876c3008f628d9
+size 65674128

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<image_soft_token>": 262144
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/README.md

@@ -0,0 +1,209 @@
+---
+base_model: google/gemma-3-4b-it
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-3-4b-it
+- 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 0.19.1

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-3-4b-it",
+  "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": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "v_proj",
+    "gate_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2270e687dd80ac694c025ba1e0911df9099255efe84b8bfa30ccb3a150773009
+size 65674128

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<image_soft_token>": 262144
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/preprocessor_config.json

@@ -0,0 +1,29 @@
+{
+  "do_convert_rgb": null,
+  "do_normalize": true,
+  "do_pan_and_scan": null,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "image_processor_type": "Gemma3ImageProcessor",
+  "image_seq_length": 256,
+  "image_std": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "pan_and_scan_max_num_crops": null,
+  "pan_and_scan_min_crop_size": null,
+  "pan_and_scan_min_ratio_to_activate": null,
+  "processor_class": "Gemma3Processor",
+  "resample": 2,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 896,
+    "width": 896
+  }
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/processor_config.json

@@ -0,0 +1,4 @@
+{
+  "image_seq_length": 256,
+  "processor_class": "Gemma3Processor"
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/special_tokens_map.json

@@ -0,0 +1,33 @@
+{
+  "boi_token": "<start_of_image>",
+  "bos_token": {
+    "content": "<bos>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eoi_token": "<end_of_image>",
+  "eos_token": {
+    "content": "<eos>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "image_token": "<image_soft_token>",
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4667f2089529e8e7657cfb6d1c19910ae71ff5f28aa7ab2ff2763330affad795
+size 33384568

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1299c11d7cf632ef3b4e11937501358ada021bbdf7c47638d13c0ee982f2e79c
+size 4689074

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-10000/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b57739a1d62581ece6188330f0721c60eb3b295a43a0bf8a9bbb5e86af5bc3c6
+size 6417

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/README.md

@@ -0,0 +1,209 @@
+---
+base_model: google/gemma-3-4b-it
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-3-4b-it
+- 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 0.19.1

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-3-4b-it",
+  "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": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "v_proj",
+    "gate_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df4ca8cf6b66c1f444658791740ffc27e9fe5b38bd5e6b8427116af5978750f7
+size 65674128

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<image_soft_token>": 262144
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/preprocessor_config.json

@@ -0,0 +1,29 @@
+{
+  "do_convert_rgb": null,
+  "do_normalize": true,
+  "do_pan_and_scan": null,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "image_processor_type": "Gemma3ImageProcessor",
+  "image_seq_length": 256,
+  "image_std": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "pan_and_scan_max_num_crops": null,
+  "pan_and_scan_min_crop_size": null,
+  "pan_and_scan_min_ratio_to_activate": null,
+  "processor_class": "Gemma3Processor",
+  "resample": 2,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 896,
+    "width": 896
+  }
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/processor_config.json

@@ -0,0 +1,4 @@
+{
+  "image_seq_length": 256,
+  "processor_class": "Gemma3Processor"
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/special_tokens_map.json

@@ -0,0 +1,33 @@
+{
+  "boi_token": "<start_of_image>",
+  "bos_token": {
+    "content": "<bos>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eoi_token": "<end_of_image>",
+  "eos_token": {
+    "content": "<eos>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "image_token": "<image_soft_token>",
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4667f2089529e8e7657cfb6d1c19910ae71ff5f28aa7ab2ff2763330affad795
+size 33384568

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1299c11d7cf632ef3b4e11937501358ada021bbdf7c47638d13c0ee982f2e79c
+size 4689074

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-11250/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b57739a1d62581ece6188330f0721c60eb3b295a43a0bf8a9bbb5e86af5bc3c6
+size 6417

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/README.md

@@ -0,0 +1,209 @@
+---
+base_model: google/gemma-3-4b-it
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-3-4b-it
+- 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 0.19.1

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-3-4b-it",
+  "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": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "v_proj",
+    "gate_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:287a53bb0cf8fa8a58fb287ead5a18909feac3c03c0139e99710f682b6b1f375
+size 65674128

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<image_soft_token>": 262144
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/preprocessor_config.json

@@ -0,0 +1,29 @@
+{
+  "do_convert_rgb": null,
+  "do_normalize": true,
+  "do_pan_and_scan": null,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "image_processor_type": "Gemma3ImageProcessor",
+  "image_seq_length": 256,
+  "image_std": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "pan_and_scan_max_num_crops": null,
+  "pan_and_scan_min_crop_size": null,
+  "pan_and_scan_min_ratio_to_activate": null,
+  "processor_class": "Gemma3Processor",
+  "resample": 2,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 896,
+    "width": 896
+  }
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/processor_config.json

@@ -0,0 +1,4 @@
+{
+  "image_seq_length": 256,
+  "processor_class": "Gemma3Processor"
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/special_tokens_map.json

@@ -0,0 +1,33 @@
+{
+  "boi_token": "<start_of_image>",
+  "bos_token": {
+    "content": "<bos>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eoi_token": "<end_of_image>",
+  "eos_token": {
+    "content": "<eos>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "image_token": "<image_soft_token>",
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4667f2089529e8e7657cfb6d1c19910ae71ff5f28aa7ab2ff2763330affad795
+size 33384568

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1299c11d7cf632ef3b4e11937501358ada021bbdf7c47638d13c0ee982f2e79c
+size 4689074

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/trainer_state.json

@@ -0,0 +1,1284 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 1250,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 0.6695285618305207,
+      "epoch": 0.008,
+      "grad_norm": 14.31633472442627,
+      "learning_rate": 1.44e-06,
+      "loss": 1.2223,
+      "mean_token_accuracy": 0.7378568470478057,
+      "num_tokens": 9922.0,
+      "step": 10
+    },
+    {
+      "entropy": 0.6495137870311737,
+      "epoch": 0.016,
+      "grad_norm": 14.104403495788574,
+      "learning_rate": 3.04e-06,
+      "loss": 1.2427,
+      "mean_token_accuracy": 0.7459693193435669,
+      "num_tokens": 19973.0,
+      "step": 20
+    },
+    {
+      "entropy": 0.6639330208301544,
+      "epoch": 0.024,
+      "grad_norm": 13.883268356323242,
+      "learning_rate": 4.64e-06,
+      "loss": 1.0494,
+      "mean_token_accuracy": 0.7489401340484619,
+      "num_tokens": 30134.0,
+      "step": 30
+    },
+    {
+      "entropy": 0.6601148784160614,
+      "epoch": 0.032,
+      "grad_norm": 7.638053894042969,
+      "learning_rate": 6.24e-06,
+      "loss": 0.6835,
+      "mean_token_accuracy": 0.7916789829730988,
+      "num_tokens": 40050.0,
+      "step": 40
+    },
+    {
+      "entropy": 0.6949334561824798,
+      "epoch": 0.04,
+      "grad_norm": 7.800208568572998,
+      "learning_rate": 7.84e-06,
+      "loss": 0.6064,
+      "mean_token_accuracy": 0.7952605724334717,
+      "num_tokens": 50153.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.76209916472435,
+      "epoch": 0.048,
+      "grad_norm": 7.195173740386963,
+      "learning_rate": 9.44e-06,
+      "loss": 0.6438,
+      "mean_token_accuracy": 0.7784603774547577,
+      "num_tokens": 60428.0,
+      "step": 60
+    },
+    {
+      "entropy": 0.7352613985538483,
+      "epoch": 0.056,
+      "grad_norm": 5.307807445526123,
+      "learning_rate": 1.1040000000000001e-05,
+      "loss": 0.5891,
+      "mean_token_accuracy": 0.796974366903305,
+      "num_tokens": 70251.0,
+      "step": 70
+    },
+    {
+      "entropy": 0.7374417603015899,
+      "epoch": 0.064,
+      "grad_norm": 3.8352510929107666,
+      "learning_rate": 1.2640000000000003e-05,
+      "loss": 0.5236,
+      "mean_token_accuracy": 0.8155859649181366,
+      "num_tokens": 80413.0,
+      "step": 80
+    },
+    {
+      "entropy": 0.7411717772483826,
+      "epoch": 0.072,
+      "grad_norm": 6.424187183380127,
+      "learning_rate": 1.4240000000000001e-05,
+      "loss": 0.5622,
+      "mean_token_accuracy": 0.8019270122051239,
+      "num_tokens": 90532.0,
+      "step": 90
+    },
+    {
+      "entropy": 0.756740826368332,
+      "epoch": 0.08,
+      "grad_norm": 5.353766441345215,
+      "learning_rate": 1.584e-05,
+      "loss": 0.5791,
+      "mean_token_accuracy": 0.790449982881546,
+      "num_tokens": 100585.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.7147179067134857,
+      "epoch": 0.088,
+      "grad_norm": 5.385560512542725,
+      "learning_rate": 1.7440000000000002e-05,
+      "loss": 0.4911,
+      "mean_token_accuracy": 0.8302230596542358,
+      "num_tokens": 110502.0,
+      "step": 110
+    },
+    {
+      "entropy": 0.7470740258693696,
+      "epoch": 0.096,
+      "grad_norm": 6.4805803298950195,
+      "learning_rate": 1.904e-05,
+      "loss": 0.5527,
+      "mean_token_accuracy": 0.8000347971916199,
+      "num_tokens": 120380.0,
+      "step": 120
+    },
+    {
+      "entropy": 0.7613708913326264,
+      "epoch": 0.104,
+      "grad_norm": 4.770519256591797,
+      "learning_rate": 2.0640000000000002e-05,
+      "loss": 0.5072,
+      "mean_token_accuracy": 0.8199988007545471,
+      "num_tokens": 130332.0,
+      "step": 130
+    },
+    {
+      "entropy": 0.7286979019641876,
+      "epoch": 0.112,
+      "grad_norm": 4.616223335266113,
+      "learning_rate": 2.224e-05,
+      "loss": 0.4932,
+      "mean_token_accuracy": 0.8223420560359955,
+      "num_tokens": 140241.0,
+      "step": 140
+    },
+    {
+      "entropy": 0.7298413753509522,
+      "epoch": 0.12,
+      "grad_norm": 4.856531143188477,
+      "learning_rate": 2.3840000000000002e-05,
+      "loss": 0.4858,
+      "mean_token_accuracy": 0.8258512139320373,
+      "num_tokens": 150311.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.759659868478775,
+      "epoch": 0.128,
+      "grad_norm": 3.717965602874756,
+      "learning_rate": 2.5440000000000004e-05,
+      "loss": 0.4982,
+      "mean_token_accuracy": 0.8219776034355164,
+      "num_tokens": 160264.0,
+      "step": 160
+    },
+    {
+      "entropy": 0.7180661201477051,
+      "epoch": 0.136,
+      "grad_norm": 4.503204345703125,
+      "learning_rate": 2.704e-05,
+      "loss": 0.4345,
+      "mean_token_accuracy": 0.8364826142787933,
+      "num_tokens": 170438.0,
+      "step": 170
+    },
+    {
+      "entropy": 0.7359680116176606,
+      "epoch": 0.144,
+      "grad_norm": 6.178208351135254,
+      "learning_rate": 2.864e-05,
+      "loss": 0.4701,
+      "mean_token_accuracy": 0.8310965120792388,
+      "num_tokens": 180524.0,
+      "step": 180
+    },
+    {
+      "entropy": 0.7207672715187072,
+      "epoch": 0.152,
+      "grad_norm": 4.033023834228516,
+      "learning_rate": 3.0240000000000002e-05,
+      "loss": 0.4388,
+      "mean_token_accuracy": 0.8398010492324829,
+      "num_tokens": 190286.0,
+      "step": 190
+    },
+    {
+      "entropy": 0.7252123236656189,
+      "epoch": 0.16,
+      "grad_norm": 4.537003517150879,
+      "learning_rate": 3.184e-05,
+      "loss": 0.4265,
+      "mean_token_accuracy": 0.8499591410160064,
+      "num_tokens": 200398.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.7307703971862793,
+      "epoch": 0.168,
+      "grad_norm": 5.052333831787109,
+      "learning_rate": 3.344e-05,
+      "loss": 0.4732,
+      "mean_token_accuracy": 0.821498817205429,
+      "num_tokens": 210449.0,
+      "step": 210
+    },
+    {
+      "entropy": 0.7535093545913696,
+      "epoch": 0.176,
+      "grad_norm": 3.8205010890960693,
+      "learning_rate": 3.504e-05,
+      "loss": 0.4588,
+      "mean_token_accuracy": 0.8302142798900605,
+      "num_tokens": 220279.0,
+      "step": 220
+    },
+    {
+      "entropy": 0.7212762236595154,
+      "epoch": 0.184,
+      "grad_norm": 5.160033226013184,
+      "learning_rate": 3.664e-05,
+      "loss": 0.4276,
+      "mean_token_accuracy": 0.8424481689929962,
+      "num_tokens": 230326.0,
+      "step": 230
+    },
+    {
+      "entropy": 0.7004913449287414,
+      "epoch": 0.192,
+      "grad_norm": 5.003087997436523,
+      "learning_rate": 3.8240000000000007e-05,
+      "loss": 0.3798,
+      "mean_token_accuracy": 0.8604909420013428,
+      "num_tokens": 240352.0,
+      "step": 240
+    },
+    {
+      "entropy": 0.6996567130088807,
+      "epoch": 0.2,
+      "grad_norm": 4.287586212158203,
+      "learning_rate": 3.984e-05,
+      "loss": 0.385,
+      "mean_token_accuracy": 0.8586375415325165,
+      "num_tokens": 250554.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7321690201759339,
+      "epoch": 0.208,
+      "grad_norm": 4.065546035766602,
+      "learning_rate": 4.144e-05,
+      "loss": 0.406,
+      "mean_token_accuracy": 0.8472627520561218,
+      "num_tokens": 260613.0,
+      "step": 260
+    },
+    {
+      "entropy": 0.7161136329174042,
+      "epoch": 0.216,
+      "grad_norm": 5.161293029785156,
+      "learning_rate": 4.304e-05,
+      "loss": 0.4411,
+      "mean_token_accuracy": 0.8383095562458038,
+      "num_tokens": 270752.0,
+      "step": 270
+    },
+    {
+      "entropy": 0.7462312340736389,
+      "epoch": 0.224,
+      "grad_norm": 3.4529647827148438,
+      "learning_rate": 4.4640000000000006e-05,
+      "loss": 0.4349,
+      "mean_token_accuracy": 0.8378861784934998,
+      "num_tokens": 280749.0,
+      "step": 280
+    },
+    {
+      "entropy": 0.7058145105838776,
+      "epoch": 0.232,
+      "grad_norm": 3.7977020740509033,
+      "learning_rate": 4.624e-05,
+      "loss": 0.4001,
+      "mean_token_accuracy": 0.8555427730083466,
+      "num_tokens": 290740.0,
+      "step": 290
+    },
+    {
+      "entropy": 0.7065004706382751,
+      "epoch": 0.24,
+      "grad_norm": 3.0234885215759277,
+      "learning_rate": 4.784e-05,
+      "loss": 0.4304,
+      "mean_token_accuracy": 0.8398854076862335,
+      "num_tokens": 300704.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.719671654701233,
+      "epoch": 0.248,
+      "grad_norm": 3.467841386795044,
+      "learning_rate": 4.944e-05,
+      "loss": 0.4409,
+      "mean_token_accuracy": 0.8362361371517182,
+      "num_tokens": 310824.0,
+      "step": 310
+    },
+    {
+      "entropy": 0.6973326683044434,
+      "epoch": 0.256,
+      "grad_norm": 4.2350006103515625,
+      "learning_rate": 5.104e-05,
+      "loss": 0.3939,
+      "mean_token_accuracy": 0.8536705911159516,
+      "num_tokens": 320720.0,
+      "step": 320
+    },
+    {
+      "entropy": 0.730055433511734,
+      "epoch": 0.264,
+      "grad_norm": 3.812591552734375,
+      "learning_rate": 5.264e-05,
+      "loss": 0.3978,
+      "mean_token_accuracy": 0.8521771967411041,
+      "num_tokens": 330747.0,
+      "step": 330
+    },
+    {
+      "entropy": 0.7058379769325256,
+      "epoch": 0.272,
+      "grad_norm": 4.266584396362305,
+      "learning_rate": 5.424e-05,
+      "loss": 0.3926,
+      "mean_token_accuracy": 0.8541984319686889,
+      "num_tokens": 340760.0,
+      "step": 340
+    },
+    {
+      "entropy": 0.7316771388053894,
+      "epoch": 0.28,
+      "grad_norm": 3.9593305587768555,
+      "learning_rate": 5.584e-05,
+      "loss": 0.4387,
+      "mean_token_accuracy": 0.838571983575821,
+      "num_tokens": 350695.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7227190256118774,
+      "epoch": 0.288,
+      "grad_norm": 3.0220916271209717,
+      "learning_rate": 5.7440000000000006e-05,
+      "loss": 0.4123,
+      "mean_token_accuracy": 0.848969680070877,
+      "num_tokens": 360713.0,
+      "step": 360
+    },
+    {
+      "entropy": 0.6917089819908142,
+      "epoch": 0.296,
+      "grad_norm": 3.9197559356689453,
+      "learning_rate": 5.9040000000000004e-05,
+      "loss": 0.3733,
+      "mean_token_accuracy": 0.8610786020755767,
+      "num_tokens": 370757.0,
+      "step": 370
+    },
+    {
+      "entropy": 0.6863218784332276,
+      "epoch": 0.304,
+      "grad_norm": 4.299370765686035,
+      "learning_rate": 6.064000000000001e-05,
+      "loss": 0.3727,
+      "mean_token_accuracy": 0.8551675200462341,
+      "num_tokens": 380916.0,
+      "step": 380
+    },
+    {
+      "entropy": 0.7062383174896241,
+      "epoch": 0.312,
+      "grad_norm": 4.69020938873291,
+      "learning_rate": 6.224e-05,
+      "loss": 0.3991,
+      "mean_token_accuracy": 0.8521957516670227,
+      "num_tokens": 390977.0,
+      "step": 390
+    },
+    {
+      "entropy": 0.6997456550598145,
+      "epoch": 0.32,
+      "grad_norm": 3.5297598838806152,
+      "learning_rate": 6.384e-05,
+      "loss": 0.4139,
+      "mean_token_accuracy": 0.845986521244049,
+      "num_tokens": 401110.0,
+      "step": 400
+    },
+    {
+      "entropy": 0.6997593820095063,
+      "epoch": 0.328,
+      "grad_norm": 5.048018932342529,
+      "learning_rate": 6.544e-05,
+      "loss": 0.3862,
+      "mean_token_accuracy": 0.8572434246540069,
+      "num_tokens": 411064.0,
+      "step": 410
+    },
+    {
+      "entropy": 0.6738950192928315,
+      "epoch": 0.336,
+      "grad_norm": 3.254060745239258,
+      "learning_rate": 6.704000000000001e-05,
+      "loss": 0.3466,
+      "mean_token_accuracy": 0.8685573637485504,
+      "num_tokens": 421145.0,
+      "step": 420
+    },
+    {
+      "entropy": 0.7274752855300903,
+      "epoch": 0.344,
+      "grad_norm": 4.334840774536133,
+      "learning_rate": 6.864000000000001e-05,
+      "loss": 0.4213,
+      "mean_token_accuracy": 0.8459632396697998,
+      "num_tokens": 430931.0,
+      "step": 430
+    },
+    {
+      "entropy": 0.7435589671134949,
+      "epoch": 0.352,
+      "grad_norm": 6.489348411560059,
+      "learning_rate": 7.024e-05,
+      "loss": 0.4025,
+      "mean_token_accuracy": 0.8490091860294342,
+      "num_tokens": 440854.0,
+      "step": 440
+    },
+    {
+      "entropy": 0.7177834689617157,
+      "epoch": 0.36,
+      "grad_norm": 4.771054744720459,
+      "learning_rate": 7.184e-05,
+      "loss": 0.3645,
+      "mean_token_accuracy": 0.8636941611766815,
+      "num_tokens": 450778.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.7479185402393341,
+      "epoch": 0.368,
+      "grad_norm": 3.6275103092193604,
+      "learning_rate": 7.344000000000002e-05,
+      "loss": 0.4165,
+      "mean_token_accuracy": 0.8424341917037964,
+      "num_tokens": 460848.0,
+      "step": 460
+    },
+    {
+      "entropy": 0.7436688542366028,
+      "epoch": 0.376,
+      "grad_norm": 3.247954845428467,
+      "learning_rate": 7.504e-05,
+      "loss": 0.382,
+      "mean_token_accuracy": 0.8507310569286346,
+      "num_tokens": 470865.0,
+      "step": 470
+    },
+    {
+      "entropy": 0.7438374400138855,
+      "epoch": 0.384,
+      "grad_norm": 3.278331756591797,
+      "learning_rate": 7.664e-05,
+      "loss": 0.3613,
+      "mean_token_accuracy": 0.8644671201705932,
+      "num_tokens": 480685.0,
+      "step": 480
+    },
+    {
+      "entropy": 0.7521087110042572,
+      "epoch": 0.392,
+      "grad_norm": 3.489924192428589,
+      "learning_rate": 7.824e-05,
+      "loss": 0.3864,
+      "mean_token_accuracy": 0.8525649666786194,
+      "num_tokens": 490635.0,
+      "step": 490
+    },
+    {
+      "entropy": 0.7417400658130646,
+      "epoch": 0.4,
+      "grad_norm": 3.8655314445495605,
+      "learning_rate": 7.984000000000001e-05,
+      "loss": 0.4267,
+      "mean_token_accuracy": 0.8403328716754913,
+      "num_tokens": 500606.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7346026539802551,
+      "epoch": 0.408,
+      "grad_norm": 3.9464962482452393,
+      "learning_rate": 8.144e-05,
+      "loss": 0.3947,
+      "mean_token_accuracy": 0.8558939337730408,
+      "num_tokens": 510542.0,
+      "step": 510
+    },
+    {
+      "entropy": 0.7398899734020233,
+      "epoch": 0.416,
+      "grad_norm": 3.2201411724090576,
+      "learning_rate": 8.304e-05,
+      "loss": 0.3788,
+      "mean_token_accuracy": 0.8501916587352752,
+      "num_tokens": 520523.0,
+      "step": 520
+    },
+    {
+      "entropy": 0.7341866433620453,
+      "epoch": 0.424,
+      "grad_norm": 5.685090065002441,
+      "learning_rate": 8.464e-05,
+      "loss": 0.4072,
+      "mean_token_accuracy": 0.8449537456035614,
+      "num_tokens": 530464.0,
+      "step": 530
+    },
+    {
+      "entropy": 0.729630422592163,
+      "epoch": 0.432,
+      "grad_norm": 3.172807455062866,
+      "learning_rate": 8.624000000000001e-05,
+      "loss": 0.3847,
+      "mean_token_accuracy": 0.8562665283679962,
+      "num_tokens": 540555.0,
+      "step": 540
+    },
+    {
+      "entropy": 0.7420088469982147,
+      "epoch": 0.44,
+      "grad_norm": 3.0816385746002197,
+      "learning_rate": 8.784e-05,
+      "loss": 0.3576,
+      "mean_token_accuracy": 0.8688248336315155,
+      "num_tokens": 550542.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.7287454545497895,
+      "epoch": 0.448,
+      "grad_norm": 3.4829037189483643,
+      "learning_rate": 8.944e-05,
+      "loss": 0.3975,
+      "mean_token_accuracy": 0.8527581334114075,
+      "num_tokens": 560556.0,
+      "step": 560
+    },
+    {
+      "entropy": 0.7329239010810852,
+      "epoch": 0.456,
+      "grad_norm": 3.681492805480957,
+      "learning_rate": 9.104000000000001e-05,
+      "loss": 0.3762,
+      "mean_token_accuracy": 0.8601427555084229,
+      "num_tokens": 570518.0,
+      "step": 570
+    },
+    {
+      "entropy": 0.7524174928665162,
+      "epoch": 0.464,
+      "grad_norm": 3.2104554176330566,
+      "learning_rate": 9.264000000000001e-05,
+      "loss": 0.379,
+      "mean_token_accuracy": 0.8545800805091858,
+      "num_tokens": 580501.0,
+      "step": 580
+    },
+    {
+      "entropy": 0.7435652434825897,
+      "epoch": 0.472,
+      "grad_norm": 4.375585079193115,
+      "learning_rate": 9.424e-05,
+      "loss": 0.3549,
+      "mean_token_accuracy": 0.868000715970993,
+      "num_tokens": 590477.0,
+      "step": 590
+    },
+    {
+      "entropy": 0.7560121715068817,
+      "epoch": 0.48,
+      "grad_norm": 3.4183433055877686,
+      "learning_rate": 9.584e-05,
+      "loss": 0.3685,
+      "mean_token_accuracy": 0.8627590477466583,
+      "num_tokens": 600446.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.7317635595798493,
+      "epoch": 0.488,
+      "grad_norm": 2.8492188453674316,
+      "learning_rate": 9.744000000000002e-05,
+      "loss": 0.3814,
+      "mean_token_accuracy": 0.8567235052585602,
+      "num_tokens": 610513.0,
+      "step": 610
+    },
+    {
+      "entropy": 0.7253688097000122,
+      "epoch": 0.496,
+      "grad_norm": 5.393974781036377,
+      "learning_rate": 9.904e-05,
+      "loss": 0.376,
+      "mean_token_accuracy": 0.8586949944496155,
+      "num_tokens": 620712.0,
+      "step": 620
+    },
+    {
+      "entropy": 0.7185450434684754,
+      "epoch": 0.504,
+      "grad_norm": 3.4982898235321045,
+      "learning_rate": 9.999997200422726e-05,
+      "loss": 0.3433,
+      "mean_token_accuracy": 0.8748471081256867,
+      "num_tokens": 630849.0,
+      "step": 630
+    },
+    {
+      "entropy": 0.7177209615707397,
+      "epoch": 0.512,
+      "grad_norm": 4.2552337646484375,
+      "learning_rate": 9.999965705214383e-05,
+      "loss": 0.3816,
+      "mean_token_accuracy": 0.8563798487186431,
+      "num_tokens": 641019.0,
+      "step": 640
+    },
+    {
+      "entropy": 0.7229529082775116,
+      "epoch": 0.52,
+      "grad_norm": 4.16381311416626,
+      "learning_rate": 9.999899215547273e-05,
+      "loss": 0.3385,
+      "mean_token_accuracy": 0.8770527482032776,
+      "num_tokens": 651190.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.7371329665184021,
+      "epoch": 0.528,
+      "grad_norm": 3.0354270935058594,
+      "learning_rate": 9.99979773188675e-05,
+      "loss": 0.3509,
+      "mean_token_accuracy": 0.8691994249820709,
+      "num_tokens": 661447.0,
+      "step": 660
+    },
+    {
+      "entropy": 0.7256860733032227,
+      "epoch": 0.536,
+      "grad_norm": 2.990354061126709,
+      "learning_rate": 9.999661254943096e-05,
+      "loss": 0.4006,
+      "mean_token_accuracy": 0.846245813369751,
+      "num_tokens": 671465.0,
+      "step": 670
+    },
+    {
+      "entropy": 0.7200690507888794,
+      "epoch": 0.544,
+      "grad_norm": 4.931247711181641,
+      "learning_rate": 9.999489785671501e-05,
+      "loss": 0.3225,
+      "mean_token_accuracy": 0.8801100552082062,
+      "num_tokens": 681438.0,
+      "step": 680
+    },
+    {
+      "entropy": 0.7339150905609131,
+      "epoch": 0.552,
+      "grad_norm": 3.662463903427124,
+      "learning_rate": 9.99928332527207e-05,
+      "loss": 0.3793,
+      "mean_token_accuracy": 0.8533829510211944,
+      "num_tokens": 691512.0,
+      "step": 690
+    },
+    {
+      "entropy": 0.7665309846401215,
+      "epoch": 0.56,
+      "grad_norm": 3.4143881797790527,
+      "learning_rate": 9.999041875189808e-05,
+      "loss": 0.3848,
+      "mean_token_accuracy": 0.8555181980133056,
+      "num_tokens": 701615.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.7422050297260284,
+      "epoch": 0.568,
+      "grad_norm": 4.814603328704834,
+      "learning_rate": 9.998765437114606e-05,
+      "loss": 0.3761,
+      "mean_token_accuracy": 0.8617783725261688,
+      "num_tokens": 711701.0,
+      "step": 710
+    },
+    {
+      "entropy": 0.7363264322280884,
+      "epoch": 0.576,
+      "grad_norm": 3.3463022708892822,
+      "learning_rate": 9.998454012981241e-05,
+      "loss": 0.3632,
+      "mean_token_accuracy": 0.8603371024131775,
+      "num_tokens": 721974.0,
+      "step": 720
+    },
+    {
+      "entropy": 0.70306516289711,
+      "epoch": 0.584,
+      "grad_norm": 5.319610118865967,
+      "learning_rate": 9.99810760496935e-05,
+      "loss": 0.3463,
+      "mean_token_accuracy": 0.8675795495510101,
+      "num_tokens": 732006.0,
+      "step": 730
+    },
+    {
+      "entropy": 0.7278270959854126,
+      "epoch": 0.592,
+      "grad_norm": 4.455011367797852,
+      "learning_rate": 9.997726215503422e-05,
+      "loss": 0.3588,
+      "mean_token_accuracy": 0.8676094055175781,
+      "num_tokens": 741957.0,
+      "step": 740
+    },
+    {
+      "entropy": 0.7247029364109039,
+      "epoch": 0.6,
+      "grad_norm": 4.075356960296631,
+      "learning_rate": 9.997309847252781e-05,
+      "loss": 0.3536,
+      "mean_token_accuracy": 0.8631299197673797,
+      "num_tokens": 751801.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.7470250844955444,
+      "epoch": 0.608,
+      "grad_norm": 3.7563424110412598,
+      "learning_rate": 9.99685850313156e-05,
+      "loss": 0.3911,
+      "mean_token_accuracy": 0.8504732131958008,
+      "num_tokens": 761804.0,
+      "step": 760
+    },
+    {
+      "entropy": 0.7228377521038055,
+      "epoch": 0.616,
+      "grad_norm": 4.742606163024902,
+      "learning_rate": 9.99637218629869e-05,
+      "loss": 0.3413,
+      "mean_token_accuracy": 0.8718061029911042,
+      "num_tokens": 771731.0,
+      "step": 770
+    },
+    {
+      "entropy": 0.723927891254425,
+      "epoch": 0.624,
+      "grad_norm": 3.5365488529205322,
+      "learning_rate": 9.995850900157875e-05,
+      "loss": 0.3776,
+      "mean_token_accuracy": 0.8463309288024903,
+      "num_tokens": 781833.0,
+      "step": 780
+    },
+    {
+      "entropy": 0.7295382738113403,
+      "epoch": 0.632,
+      "grad_norm": 3.5046541690826416,
+      "learning_rate": 9.995294648357565e-05,
+      "loss": 0.3552,
+      "mean_token_accuracy": 0.8618111193180085,
+      "num_tokens": 791888.0,
+      "step": 790
+    },
+    {
+      "entropy": 0.7007244110107422,
+      "epoch": 0.64,
+      "grad_norm": 5.335765361785889,
+      "learning_rate": 9.99470343479093e-05,
+      "loss": 0.3291,
+      "mean_token_accuracy": 0.8730599522590637,
+      "num_tokens": 801833.0,
+      "step": 800
+    },
+    {
+      "entropy": 0.7444432735443115,
+      "epoch": 0.648,
+      "grad_norm": 3.666486978530884,
+      "learning_rate": 9.994077263595842e-05,
+      "loss": 0.3524,
+      "mean_token_accuracy": 0.8662495076656341,
+      "num_tokens": 812036.0,
+      "step": 810
+    },
+    {
+      "entropy": 0.7175344467163086,
+      "epoch": 0.656,
+      "grad_norm": 3.7191364765167236,
+      "learning_rate": 9.993416139154834e-05,
+      "loss": 0.3222,
+      "mean_token_accuracy": 0.8765709340572357,
+      "num_tokens": 822052.0,
+      "step": 820
+    },
+    {
+      "entropy": 0.7220284402370453,
+      "epoch": 0.664,
+      "grad_norm": 4.217904567718506,
+      "learning_rate": 9.992720066095074e-05,
+      "loss": 0.3436,
+      "mean_token_accuracy": 0.8713880062103272,
+      "num_tokens": 832034.0,
+      "step": 830
+    },
+    {
+      "entropy": 0.702842253446579,
+      "epoch": 0.672,
+      "grad_norm": 3.448692798614502,
+      "learning_rate": 9.99198904928834e-05,
+      "loss": 0.3372,
+      "mean_token_accuracy": 0.8770883083343506,
+      "num_tokens": 842102.0,
+      "step": 840
+    },
+    {
+      "entropy": 0.7189131140708923,
+      "epoch": 0.68,
+      "grad_norm": 3.283367872238159,
+      "learning_rate": 9.99122309385097e-05,
+      "loss": 0.3282,
+      "mean_token_accuracy": 0.8700813591480255,
+      "num_tokens": 852144.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.7190994381904602,
+      "epoch": 0.688,
+      "grad_norm": 4.34605598449707,
+      "learning_rate": 9.990422205143842e-05,
+      "loss": 0.3501,
+      "mean_token_accuracy": 0.8619783520698547,
+      "num_tokens": 862145.0,
+      "step": 860
+    },
+    {
+      "entropy": 0.7451982796192169,
+      "epoch": 0.696,
+      "grad_norm": 3.9613375663757324,
+      "learning_rate": 9.989586388772327e-05,
+      "loss": 0.3651,
+      "mean_token_accuracy": 0.8604841291904449,
+      "num_tokens": 872398.0,
+      "step": 870
+    },
+    {
+      "entropy": 0.7192246913909912,
+      "epoch": 0.704,
+      "grad_norm": 5.184381484985352,
+      "learning_rate": 9.988715650586255e-05,
+      "loss": 0.3339,
+      "mean_token_accuracy": 0.8775500416755676,
+      "num_tokens": 882432.0,
+      "step": 880
+    },
+    {
+      "entropy": 0.7013565480709076,
+      "epoch": 0.712,
+      "grad_norm": 3.900893211364746,
+      "learning_rate": 9.987809996679868e-05,
+      "loss": 0.3402,
+      "mean_token_accuracy": 0.8683467149734497,
+      "num_tokens": 892404.0,
+      "step": 890
+    },
+    {
+      "entropy": 0.7010218977928162,
+      "epoch": 0.72,
+      "grad_norm": 4.807793617248535,
+      "learning_rate": 9.986869433391786e-05,
+      "loss": 0.3176,
+      "mean_token_accuracy": 0.8841704070568085,
+      "num_tokens": 902498.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.6755616426467895,
+      "epoch": 0.728,
+      "grad_norm": 3.8163652420043945,
+      "learning_rate": 9.985893967304953e-05,
+      "loss": 0.3269,
+      "mean_token_accuracy": 0.8692748427391053,
+      "num_tokens": 912732.0,
+      "step": 910
+    },
+    {
+      "entropy": 0.6885362267494202,
+      "epoch": 0.736,
+      "grad_norm": 3.0967180728912354,
+      "learning_rate": 9.984883605246596e-05,
+      "loss": 0.3229,
+      "mean_token_accuracy": 0.8779194414615631,
+      "num_tokens": 922863.0,
+      "step": 920
+    },
+    {
+      "entropy": 0.7481678783893585,
+      "epoch": 0.744,
+      "grad_norm": 2.919764280319214,
+      "learning_rate": 9.983838354288181e-05,
+      "loss": 0.3502,
+      "mean_token_accuracy": 0.8690755426883697,
+      "num_tokens": 932857.0,
+      "step": 930
+    },
+    {
+      "entropy": 0.7078751921653748,
+      "epoch": 0.752,
+      "grad_norm": 4.200019359588623,
+      "learning_rate": 9.982758221745355e-05,
+      "loss": 0.3386,
+      "mean_token_accuracy": 0.8685524761676788,
+      "num_tokens": 942814.0,
+      "step": 940
+    },
+    {
+      "entropy": 0.7074415683746338,
+      "epoch": 0.76,
+      "grad_norm": 3.062725067138672,
+      "learning_rate": 9.981643215177901e-05,
+      "loss": 0.3409,
+      "mean_token_accuracy": 0.8671726942062378,
+      "num_tokens": 952688.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.718981820344925,
+      "epoch": 0.768,
+      "grad_norm": 4.258919715881348,
+      "learning_rate": 9.98049334238968e-05,
+      "loss": 0.3399,
+      "mean_token_accuracy": 0.867408150434494,
+      "num_tokens": 962506.0,
+      "step": 960
+    },
+    {
+      "entropy": 0.7256728708744049,
+      "epoch": 0.776,
+      "grad_norm": 2.8002431392669678,
+      "learning_rate": 9.979308611428588e-05,
+      "loss": 0.3367,
+      "mean_token_accuracy": 0.8669355809688568,
+      "num_tokens": 972281.0,
+      "step": 970
+    },
+    {
+      "entropy": 0.6978988409042358,
+      "epoch": 0.784,
+      "grad_norm": 4.532417297363281,
+      "learning_rate": 9.978089030586482e-05,
+      "loss": 0.3234,
+      "mean_token_accuracy": 0.8743600666522979,
+      "num_tokens": 982438.0,
+      "step": 980
+    },
+    {
+      "entropy": 0.711053341627121,
+      "epoch": 0.792,
+      "grad_norm": 3.471184492111206,
+      "learning_rate": 9.976834608399135e-05,
+      "loss": 0.3732,
+      "mean_token_accuracy": 0.8583226442337036,
+      "num_tokens": 992443.0,
+      "step": 990
+    },
+    {
+      "entropy": 0.7319515705108642,
+      "epoch": 0.8,
+      "grad_norm": 4.972551345825195,
+      "learning_rate": 9.975545353646172e-05,
+      "loss": 0.3461,
+      "mean_token_accuracy": 0.8701602160930634,
+      "num_tokens": 1002322.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6830366253852844,
+      "epoch": 0.808,
+      "grad_norm": 4.826103687286377,
+      "learning_rate": 9.974221275351012e-05,
+      "loss": 0.3345,
+      "mean_token_accuracy": 0.872558867931366,
+      "num_tokens": 1012256.0,
+      "step": 1010
+    },
+    {
+      "entropy": 0.7266723334789276,
+      "epoch": 0.816,
+      "grad_norm": 3.449570894241333,
+      "learning_rate": 9.972862382780795e-05,
+      "loss": 0.3444,
+      "mean_token_accuracy": 0.8673858284950257,
+      "num_tokens": 1022220.0,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6967022597789765,
+      "epoch": 0.824,
+      "grad_norm": 3.9861741065979004,
+      "learning_rate": 9.971468685446332e-05,
+      "loss": 0.3188,
+      "mean_token_accuracy": 0.8719709277153015,
+      "num_tokens": 1032244.0,
+      "step": 1030
+    },
+    {
+      "entropy": 0.731421309709549,
+      "epoch": 0.832,
+      "grad_norm": 5.774566173553467,
+      "learning_rate": 9.970040193102024e-05,
+      "loss": 0.3436,
+      "mean_token_accuracy": 0.8656271934509278,
+      "num_tokens": 1042334.0,
+      "step": 1040
+    },
+    {
+      "entropy": 0.7019522726535797,
+      "epoch": 0.84,
+      "grad_norm": 4.040722846984863,
+      "learning_rate": 9.968576915745807e-05,
+      "loss": 0.3266,
+      "mean_token_accuracy": 0.8769211709499359,
+      "num_tokens": 1052249.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.7154686987400055,
+      "epoch": 0.848,
+      "grad_norm": 4.47777795791626,
+      "learning_rate": 9.967078863619065e-05,
+      "loss": 0.3253,
+      "mean_token_accuracy": 0.8803964793682099,
+      "num_tokens": 1062379.0,
+      "step": 1060
+    },
+    {
+      "entropy": 0.7162931144237519,
+      "epoch": 0.856,
+      "grad_norm": 3.963560104370117,
+      "learning_rate": 9.96554604720658e-05,
+      "loss": 0.3457,
+      "mean_token_accuracy": 0.8639598965644837,
+      "num_tokens": 1072509.0,
+      "step": 1070
+    },
+    {
+      "entropy": 0.68651864528656,
+      "epoch": 0.864,
+      "grad_norm": 3.90405535697937,
+      "learning_rate": 9.963978477236437e-05,
+      "loss": 0.3213,
+      "mean_token_accuracy": 0.8799140512943268,
+      "num_tokens": 1082525.0,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6841809570789337,
+      "epoch": 0.872,
+      "grad_norm": 3.777728796005249,
+      "learning_rate": 9.962376164679968e-05,
+      "loss": 0.3171,
+      "mean_token_accuracy": 0.8745035171508789,
+      "num_tokens": 1092387.0,
+      "step": 1090
+    },
+    {
+      "entropy": 0.6844796717166901,
+      "epoch": 0.88,
+      "grad_norm": 4.079926013946533,
+      "learning_rate": 9.960739120751661e-05,
+      "loss": 0.3341,
+      "mean_token_accuracy": 0.8692353010177613,
+      "num_tokens": 1102411.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.7135649085044861,
+      "epoch": 0.888,
+      "grad_norm": 3.8200302124023438,
+      "learning_rate": 9.959067356909086e-05,
+      "loss": 0.3343,
+      "mean_token_accuracy": 0.874856811761856,
+      "num_tokens": 1112492.0,
+      "step": 1110
+    },
+    {
+      "entropy": 0.6947476506233216,
+      "epoch": 0.896,
+      "grad_norm": 4.661466598510742,
+      "learning_rate": 9.957360884852817e-05,
+      "loss": 0.3214,
+      "mean_token_accuracy": 0.8752268016338348,
+      "num_tokens": 1122548.0,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6889101743698121,
+      "epoch": 0.904,
+      "grad_norm": 2.865563154220581,
+      "learning_rate": 9.955619716526355e-05,
+      "loss": 0.3283,
+      "mean_token_accuracy": 0.8767104983329773,
+      "num_tokens": 1132760.0,
+      "step": 1130
+    },
+    {
+      "entropy": 0.6894418656826019,
+      "epoch": 0.912,
+      "grad_norm": 4.179896831512451,
+      "learning_rate": 9.953843864116024e-05,
+      "loss": 0.2963,
+      "mean_token_accuracy": 0.8876428008079529,
+      "num_tokens": 1142862.0,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6653194487094879,
+      "epoch": 0.92,
+      "grad_norm": 3.7244317531585693,
+      "learning_rate": 9.952033340050914e-05,
+      "loss": 0.314,
+      "mean_token_accuracy": 0.8793320178985595,
+      "num_tokens": 1153102.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.6848842680454255,
+      "epoch": 0.928,
+      "grad_norm": 3.8479301929473877,
+      "learning_rate": 9.95018815700277e-05,
+      "loss": 0.3288,
+      "mean_token_accuracy": 0.8732554018497467,
+      "num_tokens": 1163144.0,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6730503559112548,
+      "epoch": 0.936,
+      "grad_norm": 3.0569231510162354,
+      "learning_rate": 9.948308327885921e-05,
+      "loss": 0.3076,
+      "mean_token_accuracy": 0.8854502260684967,
+      "num_tokens": 1173311.0,
+      "step": 1170
+    },
+    {
+      "entropy": 0.6516952633857727,
+      "epoch": 0.944,
+      "grad_norm": 4.482373237609863,
+      "learning_rate": 9.946393865857175e-05,
+      "loss": 0.2849,
+      "mean_token_accuracy": 0.8902148902416229,
+      "num_tokens": 1183248.0,
+      "step": 1180
+    },
+    {
+      "entropy": 0.674964702129364,
+      "epoch": 0.952,
+      "grad_norm": 3.295825242996216,
+      "learning_rate": 9.944444784315737e-05,
+      "loss": 0.3198,
+      "mean_token_accuracy": 0.880325585603714,
+      "num_tokens": 1193239.0,
+      "step": 1190
+    },
+    {
+      "entropy": 0.6798981428146362,
+      "epoch": 0.96,
+      "grad_norm": 3.836064577102661,
+      "learning_rate": 9.942461096903111e-05,
+      "loss": 0.3209,
+      "mean_token_accuracy": 0.876353645324707,
+      "num_tokens": 1203264.0,
+      "step": 1200
+    },
+    {
+      "entropy": 0.7014159679412841,
+      "epoch": 0.968,
+      "grad_norm": 3.5167601108551025,
+      "learning_rate": 9.940442817503006e-05,
+      "loss": 0.3219,
+      "mean_token_accuracy": 0.870438277721405,
+      "num_tokens": 1213358.0,
+      "step": 1210
+    },
+    {
+      "entropy": 0.6847694993019104,
+      "epoch": 0.976,
+      "grad_norm": 3.9210779666900635,
+      "learning_rate": 9.938389960241237e-05,
+      "loss": 0.3163,
+      "mean_token_accuracy": 0.8697898685932159,
+      "num_tokens": 1223432.0,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6633544981479644,
+      "epoch": 0.984,
+      "grad_norm": 4.78698205947876,
+      "learning_rate": 9.93630253948563e-05,
+      "loss": 0.3127,
+      "mean_token_accuracy": 0.8793421804904937,
+      "num_tokens": 1233517.0,
+      "step": 1230
+    },
+    {
+      "entropy": 0.6780155301094055,
+      "epoch": 0.992,
+      "grad_norm": 3.831955671310425,
+      "learning_rate": 9.934180569845917e-05,
+      "loss": 0.3139,
+      "mean_token_accuracy": 0.8776540398597718,
+      "num_tokens": 1243535.0,
+      "step": 1240
+    },
+    {
+      "entropy": 0.6700406908988953,
+      "epoch": 1.0,
+      "grad_norm": 4.903558731079102,
+      "learning_rate": 9.932024066173635e-05,
+      "loss": 0.3127,
+      "mean_token_accuracy": 0.8813739597797394,
+      "num_tokens": 1253644.0,
+      "step": 1250
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 12500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.10205773753344e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-1250/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b57739a1d62581ece6188330f0721c60eb3b295a43a0bf8a9bbb5e86af5bc3c6
+size 6417

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-12500/README.md

@@ -0,0 +1,209 @@
+---
+base_model: google/gemma-3-4b-it
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-3-4b-it
+- 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 0.19.1

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-12500/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-3-4b-it",
+  "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": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "v_proj",
+    "gate_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-12500/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e396b5c0227fc1a95d5ec2a972d9d020700c32ca7a642ae3bc876c3008f628d9
+size 65674128

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-12500/added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<image_soft_token>": 262144
+}

checkpoints/otter_gemma3_4b_r8_a32_adamw_e10_lr1e-4_s2_baseline/checkpoint-12500/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}